Steel ball sorting based on electromagnetic induction using eddy current technique

Abstract

There were many machine failures caused by the damage of the rolling bearing, for which the main reason was the quality of the steel ball [1]. Many works have been reported on assessing the geometric quality of steel balls, such as shape, size, and surface defects. In contrast, testing the material properties of steel balls is rarely mentioned [2]. As the quality and working life of the bearing depend primarily on the material properties of the steel ball, it is necessary to unambiguously classify the properties of steel balls in making bearings and other applications to ensure the reliability of the mechanical machine system. In this digest, we report on the eddy-current technique for the classification of steel balls based on the amplitude and phase of the eddy-current signal. Three types of 4-mm steel balls, including carbon, chrome, and stainless 304 steel balls, were tested to determine if it had undergone the surface hardening process.Fig. 1 shows the components of the RLC-resonance eddy-current system, which was designed to convert the electromagnetic induction of the steel ball into the phase-sensitive voltage output signals. The pickup coil is mounted statically to detect the eddy-current signal when the steel ball moves back and forth through the pickup coil by the stepping motor translation stage. A microcontroller (Arduino Uno) is used as the main unit to gather and process the data of the stage, such as the position and velocity. The micro-step determined on the driver is set to be 1/32.Fig. 1. The experimental eddy-current system for steel ball sorting.To detect the signal, the reference waveform for the lock-amplifier is synchronized to the excitation signal of the eddy current sensor. The lock-in amplifier receives the output signal from the sensor and generates the analog output voltages in proportion to the in-phase (real) and quadrature-phase (imaginary) components of the signal. The voltages are digitized by the data acquisition (DAQ) device NI USB 6008 from the National Instruments. The translation stage and the DAQ device are both controlled by the computer with the graphic user interface software designed using the Visual Studio C# programming language.The real and imaginary voltage components of the recorded eddy-current signals are subsequently converted into the probability map to show whether the steel balls can be classified by eddy-current sensors. The change in the detected signal should be due to the difference in the electromagnetic properties of the steel balls since the ball diameters are all 4 mm with an error less than 10 μm.Fig. 2. Eddy-current classification of steel balls: (a) determining the resonant frequency, (b) finding out the optimal resistance, and (c) voltage probability map of the steel balls before and after the surface hardening process.The impendence of the pickup coil can be calculated according to Dodd and Deeds method [3]:Zc=[(2πωµn2rm)/(l2-l1)2(r2-r1)2]∫I2[(r2,r1)/α5]{(l2-l1)(1+e-α/α)-1/α}dα (1)where Zc is the impedance of the coil (Ohm), µ is the permeability (Henry/meter), I is the applied current (ampere), α is the separation constant (meter-1), r2 is the outer radius of the coil (meter), r1 is the inner radius of the coil (meter), rm is the average radius of the coil, ω is the angular frequency (rad/second), l2 is the distance from metal testing to top of the coil (meter), l1 is the distance from metal testing to bottom of the coil (meter), n is the number of turns (turns), respectively. Fig. 2(a) shows the voltage amplitude measured at the output terminal when the driving frequency varies from 50 kHz to 200 kHz. The output response is maximized at the resonant frequency of 116 kHz for the RLC circuit. To achieve the higher response, the voltage outputs with various resistors are measured, as shown in Fig.2(b), where the optimal resistance is found to be 20 kΩ. The demodulated eddy-current signals were collected and repeated 100 times for each type of ball, and the result is converted into the voltage probability map in Fig. 2(c). It is found that the steel balls before and after the surface hardening treatment can be classified based on the difference in the eddy-current signal caused by the variation in electrical conductivity and permeability. The result indicates that the proposed method is feasible for application in sorting the material quality of steel balls, which plays a crucial role in mechanical systems.The authors would like to thank Tan Kong Precision Tech Co. Ltd. for providing the steel ball samples. This work is supported by the Ministry of Science and Technology of Taiwan under Grant No. MOST108-2221-E992-083MY2. **